Asymptotic analysis - if f(n) = Ω(g(n)), how to prove ln(f(n)) = Ω(ln(g(n)))?

2k Views Asked by At

Is the following statement true, if so, how can I prove it?

if f(n) = Ω(g(n)), is also true that ln(f(n)) = Ω(ln(g(n)))?

Since f(n) >= c*g(n), I divided the problem into two branches:

If f(n) = g(n), it's also true that ln(x) = ln(x) and the statement it's true.

If f(n) > g(n), then the value for ln is greater, if the argument is greater:

x > y; ln(x) > ln(y)

Is there something wrong with my reasoning?

1

There are 1 best solutions below

4
On BEST ANSWER

This is false in general without some further conditions.

For example take $f(n) = n$ and $g(n) = 2n$. Then $f(n) \geq \frac{1}{2} g(n)$ on the domain $n \geq 1$ but $\log f(n) \not\geq c \log g(n)$ on the domain $n \geq 1$ for any $c > 0$. In this case it is only true that $\log f(n) = \Omega(\log g(n))$ as $n \to \infty$.

For another example take $f(n) = \sin(n\pi/6) + 2$ and $g(n) = \sin(n\pi/6) + 3$. Then $f(n) \geq \frac{1}{2} g(n)$ for all $n$ but $\log f(n) \not\geq c \log g(n)$ for any $c > 0$ on any domain that includes the points $n=9,21,33,45,\ldots$

The statement is true if $g(n) \to \infty$ as $n \to \infty$ and $f(n) = \Omega(g(n))$ as $n \to \infty$. For then we can find a $c > 0$ such that

$$ f(n) \geq c\,g(n) $$

for all $n$ sufficiently large, from which

$$ \log f(n) \geq \log c + \log g(n) \geq \frac{1}{2}\log g(n) $$

for $n$ large enough. Thus $\log f(n) = \Omega(\log g(n))$ as $n \to \infty$.